Today's Image & Video Generation: Fastest-Growing Projects — October 25, 2026
Today's the Image & Video Generation space on GitHub, there's a noticeable trend towards AI-driven creation and enhancement of visual content, with an emphasis on both artistic expression and technical innovation. The manga-generator tool stands out as one such example by facilitating the end-to-end production of manga-style comics through scriptwriting to PDF output.
The stockedge/manga-generator repository has seen significant growth this week with a Growth Score of 33.83 and 74 stars, reflecting its utility for those interested in creating or automating parts of the comic creation process using AI. It offers an integrated workflow from script writing through image generation to final PDF output.
Yangsir2077's pixel-loom is an open-source tool that supports AI-driven pixel art and animation creation via MCP (Minecraft Protocol). With a Growth Score of 15.15 and 87 stars, it caters to artists who want to leverage AI in the creative process for pixel-based artwork and animations.
Isixe's bg2x is a privacy-focused AI tool designed for image super-resolution or upscaling. It has garnered 133 stars and a Growth Score of 13.36, indicating its relevance for users looking to enhance the resolution of images without compromising on quality or privacy concerns.
Speridlabs' iris-3b project is notable for its innovative approach to pixel-space generation and general vision learning. This diffusion transformer directly generates every pixel without using a Variational Autoencoder (VAE), making it versatile for tasks like monocular depth estimation and image restoration. With 204 stars and a Growth Score of 8.40, iris-3b demonstrates the potential of AI in various visual tasks beyond simple upscaling.
SimonGiebenhain's video-diffusion-avatars repository is dedicated to transforming video diffusion models into controllable 3D avatars. Although it has a lower Growth Score of 5.17 and 57 stars, its specific focus on creating interactive 3D characters from video data sets it apart in the realm of generative AI for animation and gaming.
Harshpreet931's harshpreet-hand showcases an interesting use case where Claude Code was fed handwritten notes to generate a personalized font. This project has gained attention with 53 stars and a Growth Score of 3.88, highlighting how AI can be used creatively to convert personal handwriting into digital fonts.
PolyU-VCLab's HarnessIR is another noteworthy project aimed at harnessing multimodal foundation models for real-world image restoration tasks. With 16 commits in the last month and 34 stars, it reflects the growing interest in using advanced models for practical image enhancement applications.
VectorSpaceLab's InfoVis repository stands out with its ambitious goal of creating a single model capable of generating and editing slide images alongside a benchmarking framework. Its Growth Score of 2.78 and 41 stars suggest steady growth as researchers and developers integrate it into their workflows.
Worldbench's Orbia project explores the consistency of generated video content in forming a coherent 3D world, which is crucial for realistic virtual environments. With 6 commits in the last month and 49 stars, its Growth Score of 2.50 indicates interest from those working on advanced video generation techniques.
Lastly, sunshy-1's StyleForge project investigates eliciting inherent style transfer capabilities within step-distilled diffusion models, aiming to refine how AI can manipulate visual styles effectively. With a Growth Score of 2.07 and 27 stars, it is an emerging tool for developers interested in stylistic transformations in image generation.
Overall, these projects highlight the diverse applications of AI in generating and enhancing images and videos, from creative arts to technical image restoration and beyond.
The stockedge/manga-generator repository has seen significant growth this week with a Growth Score of 33.83 and 74 stars, reflecting its utility for those interested in creating or automating parts of the comic creation process using AI. It offers an integrated workflow from script writing through image generation to final PDF output.
Yangsir2077's pixel-loom is an open-source tool that supports AI-driven pixel art and animation creation via MCP (Minecraft Protocol). With a Growth Score of 15.15 and 87 stars, it caters to artists who want to leverage AI in the creative process for pixel-based artwork and animations.
Isixe's bg2x is a privacy-focused AI tool designed for image super-resolution or upscaling. It has garnered 133 stars and a Growth Score of 13.36, indicating its relevance for users looking to enhance the resolution of images without compromising on quality or privacy concerns.
Speridlabs' iris-3b project is notable for its innovative approach to pixel-space generation and general vision learning. This diffusion transformer directly generates every pixel without using a Variational Autoencoder (VAE), making it versatile for tasks like monocular depth estimation and image restoration. With 204 stars and a Growth Score of 8.40, iris-3b demonstrates the potential of AI in various visual tasks beyond simple upscaling.
SimonGiebenhain's video-diffusion-avatars repository is dedicated to transforming video diffusion models into controllable 3D avatars. Although it has a lower Growth Score of 5.17 and 57 stars, its specific focus on creating interactive 3D characters from video data sets it apart in the realm of generative AI for animation and gaming.
Harshpreet931's harshpreet-hand showcases an interesting use case where Claude Code was fed handwritten notes to generate a personalized font. This project has gained attention with 53 stars and a Growth Score of 3.88, highlighting how AI can be used creatively to convert personal handwriting into digital fonts.
PolyU-VCLab's HarnessIR is another noteworthy project aimed at harnessing multimodal foundation models for real-world image restoration tasks. With 16 commits in the last month and 34 stars, it reflects the growing interest in using advanced models for practical image enhancement applications.
VectorSpaceLab's InfoVis repository stands out with its ambitious goal of creating a single model capable of generating and editing slide images alongside a benchmarking framework. Its Growth Score of 2.78 and 41 stars suggest steady growth as researchers and developers integrate it into their workflows.
Worldbench's Orbia project explores the consistency of generated video content in forming a coherent 3D world, which is crucial for realistic virtual environments. With 6 commits in the last month and 49 stars, its Growth Score of 2.50 indicates interest from those working on advanced video generation techniques.
Lastly, sunshy-1's StyleForge project investigates eliciting inherent style transfer capabilities within step-distilled diffusion models, aiming to refine how AI can manipulate visual styles effectively. With a Growth Score of 2.07 and 27 stars, it is an emerging tool for developers interested in stylistic transformations in image generation.
Overall, these projects highlight the diverse applications of AI in generating and enhancing images and videos, from creative arts to technical image restoration and beyond.